The existence of a competency solely responsible for processing decisions involving sacred values means that what was thought to be a unitary target of inquiry, decision-making, actually contains two targets, sacred decision-making and non-sacred decision-making. Discoveries that a natural phenomenon possesses only illusory underlying unity are not uncommon in science, occurring when we find out there are four species of giraffe or identify etiologically-distinct illnesses with identical symptomology. A successful theory has a scope that is sensitive to differences between distinct types of phenomena that merely appear similar, ensuring that theoretical entities covering heterogeneous objects do not confuse the interpretation of experimental results. Importantly, the cluster of human behavior within the scope of ISVT more or less corresponds to a pre-theatrically identifiable class of actions centered around religion and politics. These considerations raise the question of how to theorize about behavior outside the scope of ISVT; namely, actions emerging from choices between two profane outcomes.
While the contents of sacred values usually concern exceptional religious and political states of affairs, non-sacred values generally apply to the mundane outcomes that occupy most of
our attention. Countless inconsequential decisions must be made on a daily basis; whether to drive to work or take the train, whether to employ one investment strategy rather than the other, or what to cook for dinner. When people are making these decisions they are often, though not always, attempting to maximize or economize a finite resource such as time, money, or effort. One considers the choice of automobile instead of light rail ‘correct’ if the time spent commuting via automobile ends up being less than the light rail—assuming the agent is not an environmentalist, whose decision will involve ISVT. Initially, it seems reasonable to preserve RCT for explaining this class of actions because agential goals align with the decision procedure’s guarantee that a rational agent will experience the greatest benefit in the long run. In fact, restricting RCT’s scope to only profane matters improves its descriptive adequacy, given that moral choices have always resisted purely rational terms. This move rules out the most ardent interpretations of RCT (which claim its decision procedure to be operative across all choices), but enables a theorist to narrow the explanatory scope to a specific cluster of actions, as was done in ISVT.
However, many scholars would justifiably question the use of RCT in light of recent interest in non-rational understandings of decision-making like prospect theory and behavioral economics more generally (Kahneman & Tversky 1979; Ainsley 2002; Gigerenzer & Selten 2002). Although I will draw the reader’s attention to both empirical and conceptual reasons to be skeptical of RCT’s demise in this section, I first wish to make clear the reasons I take on the defense of such an unpopular theory in the present context. Further, I must note that because I consider ISVT to be an isolated competency, its truth or falsehood does not depend on the status of RCT, meaning the risks to my most substantive proposals are minimal (as are the benefits, one may think). In fact, this discussion will improve our understanding of ISVT for two reasons. Firstly, RCT’s decision procedure uses formalized and idealized principles to model decision-making while
remaining agnostic towards content, which represents a similar approach to ISVT, albeit with different rules. The similarity in theoretical features enables a useful contrast between ISVT and a decision procedure with different rules and functions that processes everyday choices. Secondly, accounts of multiple types of decision-making afford insight into the shifts in behavior and demeanor across moments in which one or the other competency is operative. A person must split his attention between the social concerns of sacred values and the more individualistic concerns of material preferences depending on the context and outcomes involved.
To make good on these promises, I first will briefly present the important concepts necessary for understanding RCT, as well as some common criticisms. I then introduce Chomsky’s performance/competence distinction to characterize seemingly adverse experimental results as performance errors. Following this, I provide two positive arguments for RCT: the first focuses on the way in which agents correct performance errors upon further consideration, the second on the conceptual necessity of assuming rationality. Finally, I discuss the implications of a motivational system split between two separate competencies that produce behavior using different rules of practical reasoning.
The central contention of RCT is that, if an agent’s preferences accord with a few simple axioms, they can be described by a ‘utility function’ that ranks all possible outcomes available to the agent in terms of the amount of utility, which can be fleshed out in terms of pleasure, money, or the good. Defined in this way, preferences integrate judgments of value with beliefs about the likelihood of events, resulting in a consistent ranking of outcomes that represents the expected utility of each possible action. An agent who always performs the action with the highest expected value is ‘rational.’ A rational agent will experience more utility, in the long run, than someone who does not maximize expected utility (see von Neumann and Morgenstein (1944) for a
mathematical proof). Thus, the image of human nature motivating RCT is of people attempting to do the best they can, whatever their circumstances.
In order to guarantee the greatest reward possible, RCT imposes axiomatic restrictions on the preference set. While there are a few variants of RCT, each with slightly different axioms, none of these divergences are relevant (Jeffery 1965; Savage 1954; von Neumann & Morgenstern 1944). No matter the system, the two most important axioms are transitivity and completeness. The transitivity axiom limits an agent’s ranking of more than two options, stating: ‘if A ≻B and B
≻ C, then A ≻ C.’ This rule ensures preferences are consistent with one another, eliminating preferential cycles that make the agent liable to be money-pumped. Completeness requires that a rational agent can form a preference between any two options or is indifferent between them. More formally, the completeness axiom states: ‘given options A and B, either A ≻B, B ≻A, or A = B.’ This assumption is required to ensure that the agent experiences the greatest expected utility possible because it guarantees that no option outside the calculus has a greater utility than those included. It further “entails that any two options, no matter how disparate, can be compared,” which means an agent is able to rank situations with very different types of good (Peterson 2009). This axiom conflicts with inviolability and unrankability, since both preclude comparisons between the types of value represented in their outcomes
Despite RCT’s staying power, the approach has faced a number of challenges, most of which concern behaviors that contradict the axioms. Behavioral economics emerged from the influential work of Amos Tversky and Daniel Kahneman, who identified ‘heuristics and biases’ that cause violations of basic rules of logic, probability, and decision-making. For some examples, Tversky challenges the transitivity axiom by finding that some experimental subjects exhibited intransitivity over series of choices lasting five weeks when the options involved multiple
attributes, like selecting college applicants for admission (1969). Similarly, both Tversky and Kahneman found that differential framing of identical options will affect an experimental subject’s rankings. The famous example is the ‘Asian Disease Problem,’ which presents a choice between healthcare policies that are framed in terms of lives-saved or number of deaths. Presenting the policy in terms of deaths moved people to choose more risky policies, presumably to avoid the salient costs of death (Tversky & Kahneman 1981). Recently, George Ainsle showed we discount future rewards in a hyperbolic curve that diverges from the exponential discounting curve implied by RCT. He argues that people discount future rewards too greatly (even when the reward is quite soon), generating irrational patterns of choice like addiction and procrastination (Ainsle 2002). These three findings are representative, but certainly not exhaustive, of experiments indicating that actual behavior is irrational according to RCT’s axioms, seeming to render the theory untenable because of anomalous results.
However, there is an alternative reading of the non-rational behavior which distinguishes between competence and performance, first introduced by Noam Chomsky. A competency is a cognitive system able to perform some useful function, governed by rules that determine the output of the system (1965). A competency enables some advantageous behavior or mental processes, such as Chomsky’s universal grammar that enables the existence of language. Chomsky’s theory of grammar is an idealized system of recursive rules that structure language, capable of generating an infinite number of sentences. Likewise, RCT’s axioms describe a competence by which agents evaluate possible options and make choices in a way that ensures the most benefit. The theory presents a logic of decision-making that necessarily ignores psychological limitations for the sake of building a description that mathematically ensures the highest possible reward.
Chomsky’s grammar and RCT are both idealized theories in the sense that their full expression would assume cognitive capacities beyond ours. RCT does not introduce restrictions on the number of options available to an agent at any one time, which, for anyone, includes a tremendous number of actions. Recall, the completeness axiom dictates that every option be included in the utility function, requiring the ability to process a nearly infinite amount of information. In actuality, people could not possibly consider all available options, limiting consideration to relatively few outcomes. Chomsky’s grammar is also capable of generating more information than a person can handle, which he illustrates with unintelligible, yet grammatically correct, sentences with many nested clauses, such as, “the man who the boy who students recognized pointed out is a friend of mine” (1965 p. 11). When rules are open ended or recursive, it is impossible for limited cognitive agents to handle very long chains of options or sentences. However, despite its failure to achieve literal truth, RCT may still improve our understanding of choice by making accurate predictions and demarcating a particular class of actions with the function of ensuring an agent derives the most benefit from her circumstances.
Since Chomsky’s universal grammar and RCT are both idealized descriptions of a competency’s structure, some behavior will deviate from theoretical expectations—failures called ‘performance errors.’ Chomsky argues that there is a “fundamental distinction between competence (the speaker-hearer's knowledge of his language) and performance (the actual use of language in concrete situations)” (1965 p. 4). The theory of a competency must bracket some of actual behavior when it obscures the patterns visible without noise caused by theoretically “irrelevant conditions [such] as memory limitations, distractions, shifts of attention and interest, and errors” (Chomsky 1965 p. 3). When someone stops and starts in the middle of a sentence, incites disagreement between a subject and a verb, or leaves a fragment unfinished, we do not
think them completely ignorant of grammatical rules, but assume the particular performance was in error. Similarly, an agent who uses the representativeness heuristic is not irredeemably irrational, but commits a performance error relative to ideal expectations.
However, Chomsky mentions some causes of performance errors that do not seem to be connected to idealization, such as “shifts in attention and interest,” but to the interaction of the competency in question with other systems in the brain (Chomsky 1965 p. 3). Competencies are enmeshed in complex networks of systems with their own functions, enabling other systems to causally influence behavior. If an angry person acts contrary to rational expectations (Hursthouse 1991), it is not immediately obvious that this counts as evidence against a theory of, specifically and only, choice; had the emotional system not interfered, the agent would have acted differently. The complexity and multiplicity of connections between different cognitive systems leave cognitive scientists with no choice but to abstract away from much of it. One introduces abstraction into the theory, when necessary, by simplifying the effects of external systems or ignoring them altogether. For example, ISVT’s treatment of the emotional resonance of sacred values is highly abstract, discussed only in terms like ‘moral outrage’ that do not differentiate between specific emotions, each of which may influence decision-making uniquely.
These theories of choice, RCT and ISVT, are not meant to be models that perfectly map onto actual structures and processes in the brain that produce behaviors. Rather, they are highly idealized and abstract theories constructed for the purpose of producing accurate predictions, demarcating domains of study, and establishing coherent functional descriptions of systems that are meant to aid human conduct in some way. Although we do not live in a world, nor are we the sort of beings, that allows for perfect agreement with the stringent standards set by RCT, we are nonetheless subject to the rational norms that generally structure our behavior. Before moving on,
I wish to introduce a further source of what appear to be performance errors, which is implied by the above meta-theoretical commitments. It is a live possibility that other domains of human action exist that neither ISVT nor RCT have the conceptual resources to explain, requiring a dedicated theory. I do not think, for instance, that decisions about romantic matters are made using solely rational faculties. If distinct clusters of choice-phenomena require splitting for descriptive adequacy, theories with improper scope can expect confused experimental results.
I now consider whether or not my construal of performance errors can aid RCT in responding to criticisms grounded in experimental results that show actual behavior to be incompatible with RCT. I begin with the observation that people accept their mistakes and adjust accordingly when realizing an error in practical or statistical reasoning. The tendency to self-correct conclusively settles the dilemma between characterizing an anomalous result as a performance error or evidence of the falsehood of the theory, in favor of the former. If people are able to correct their errors without assistance, there must be a capacity that facilitates the correction in the direction that accords with rational expectations. Without a system able to perform the necessary inferences and calculations, one would expect corrective adjustments to vary randomly rather than towards rational expectations. As my two examples will show, people, without direction, modify their choices and estimations of probability to accord with rational expectations in a range of conditions. Therefore, we must posit a competency able to perform the decision procedure described by RCT to account for our corrections of performance errors. Interestingly, there is a marked contrast between the ease of corrections in RCT’s domain and the intransigence of devoted actors confronted with challenges to the propriety of their sacred actions; in the latter case, rational suggestions would be rejected out of hand because rational standards do not apply to sacred values. The acceptance of mistakes and subsequent
revisions mean that we take ourselves to be acting in accordance with RCT, even if mistakes sometimes cause deviations.
Figure 1 Hyperbolic vs. Exponential Discounting
An important application of RCT is in choices about the future, with the theory mandating a gradual reduction in value to accompany increasingly delayed gratification. If one must wait to acquire something, it is reasonable to discount its value in proportion to the amount of time one must wait. Exponential discounting curves represent the rational expectation of our evaluations of future rewards, with shorter delays reflected in the valuation of the outcome (see #1 in the graph). Recall, however, that Ainslie (2002) argues that people discount future rewards far too much until just before the good becomes available, when the values rises very steeply (#2 in the graph). Andre Hofmeyr, Don Ross, and colleagues, provide evidence that cigarette smokers are more likely to make choices in accordance with an exponential discount curve in two conditions: first, when the salience of future rewards are heightened; and second, when subjects think about their decision as part of a ‘sequence’ (2010). These revisions toward rational expectations show that hyperbolic temporal discounting is not ubiquitous and that people possess the capacity for discounting and in a rational manner. Importantly, the corrections are not deferrals to experts pointing out mistakes,
but the result of increased attention to features of the circumstances relevant to RCT’s decision procedure.
In the statistical domain, Richard Nisbett and colleagues performed experiments on gambling behavior, with manipulations that made chance more salient and induced awareness of the range of possible outcomes. In both conditions, subjects were more responsive to “statistical considerations having to do with the adequacy of the sample,” meaning they did not overgeneralize (i.e. use the representativeness heuristic (Kahneman & Tversky 1974)) to the same degree as control groups (Nisbett et. al. 1983 p.353). Once again, that one does not need explicit instruction to self-correct implies that people can act in accordance with rational prescriptions. The
willingness to self-correct, in distinction from the direction of correction, implies that people believe that they should be following rational expectations when acting.
Of course, this treatment of two studies showing that people self-correct for violations of RCT is not sufficient to invalidate many years of research on non-rational behavior. They do, however, establish two crucial propositions that must be reconciled with the results of behavioral economics: we have the capacity to act according to rational principles and we believe that we should be acting according to rational principles. In addition, despite the volume of studies that show irrationality, articles that lend credence to RCT continue to be published; for example, in the contexts of transitivity and framing effects (De Martino et. al. 2006; Regenwetter & Dana 2011). So, we are faced with a confusing set of experimental results that may include the operation of multiple competencies, performance errors that only appear to be anomalous results, and contraindications of rationality and irrationality seem to shift with context. Responding to this situation requires theories with proper scope that can represent, in an abstract and idealized manner, the functional properties of choices within the domain of the theory.
In my second argument, I will claim that the assumption of general, albeit imperfect, rationality furnishes theorists with a purposive explanation of many choices between two profane values. Intuitively, RCT’s decision procedure seems to be appropriate explanatory paradigm for a great many cases, from choices about the best route through traffic to searching for deals at the grocery. We make many purely instrumental decisions, most often without even noticing, that are accurately described as a process of weighing each outcome’s utility, considering the subjective